Upload
others
View
3
Download
0
Embed Size (px)
Citation preview
Algorithm Theoretical
Basis Document
For Cloud Amount
Code:NMSC/SCI/ATBD/CA
Issue:1.0 Date:2012.12.26
File: CA-ATBD_V4.0.hwp
Page : 1/22
National Meteorological Satellite Center
CA Algorithm Theoretical Basis Document
NMSC/SCI/ATBD/CA, Issue 1, rev.4
26 December 2012
Algorithm Theoretical
Basis Document
For Cloud Amount
Code:NMSC/SCI/ATBD/CA
Issue:1.0 Date:2012.12.26
File: CA-ATBD_V4.0.hwp
Page : 1/22
National Meteorological Satellite Center
REPORT SIGNATURE TABLE
Function Name Signature Date
Prepared by Yong-Sang Choi,
Heaje Cho 26 December 2012
Reviewed by Yong-Sang Choi 26 December 2012
Authorised by NMSC 26 December 2012
Algorithm Theoretical
Basis Document
For Cloud Amount
Code:NMSC/SCI/ATBD/CA
Issue:1.0 Date:2012.12.26
File: CA-ATBD_V4.0.hwp
Page : 1/22
National Meteorological Satellite Center
DOCUMENT CHANGE RECORD
Version Date Pages Changes Version5 2012.12.26 - -Nothing has changed for contents besides ATBD form.
Algorithm Theoretical
Basis Document
For Cloud Amount
Code:NMSC/SCI/ATBD/CA
Issue:1.0 Date:2012.12.26
File: CA-ATBD_V4.0.hwp
Page : 1/22
National Meteorological Satellite Center
Table of contents
1. Overview
2. Background and purpose
3. Algorithm
3.1 Theoretical background and basis
3.2 Retrieval method
3.2.1 Look up table retrieval method
3.3 Retrieval process
3.3.1 CA Retrieval
3.4 Validation
3.4.1 Validation method
3.4.2 Validation data
3.4.3 Temporal and spatial collocation method
3.4.4 Validation result analysis
4. Analysis method of retrieval results
5. Problems and possibilities for improvement
6. References
Algorithm Theoretical
Basis Document
For Cloud Amount
Code:NMSC/SCI/ATBD/CA
Issue:1.0 Date:2012.12.26
File: CA-ATBD_V4.0.hwp
Page : 1/22
National Meteorological Satellite Center
List of Tables
Table 1 : Cloud top pressure and cloud bottom height corresponding to cloud type.
Table 2 : Detailed output data for the CA algorithm.
Table 3 : QC flag.
Table 4 : Validation results of CF and CA
Table 5 : Detailed output data for the CA algorithm.
Table 6 : Quality test result for the CA algorithm.
Algorithm Theoretical
Basis Document
For Cloud Amount
Code:NMSC/SCI/ATBD/CA
Issue:1.0 Date:2012.12.26
File: CA-ATBD_V4.0.hwp
Page : 1/22
National Meteorological Satellite Center
List of Figures
Figure 1 : Images of (a), (b)nadir-view cloud fraction and (c), (d)fractional sky cover.
Figure 2 : Illustration of the parameters.
Figure 3 : Flowchart of CF and CA algorithm.
Figure 4 : Validation results of cloud fraction(a) and sky cover(b) for SGP site. Larger
circles in b represent improver Nhemisphvalues compared with Nnadir
Figure 5 : monthly mean bias between ground measured cloud fraction and MODIS
in the site.
retrieved cloud fraction (algorithm-retrieved sky cover).
Algorithm Theoretical
Basis Document
For Cloud Amount
Code:NMSC/SCI/ATBD/CA
Issue:1.0 Date:2012.12.26
File: CA-ATBD_V4.0.hwp
Page : 1/22
National Meteorological Satellite Center
List of Acronyms
ARM Atmospheric Radiation Measurement
CA Cloud Amount
CF Cloud Fraction
CMDPS COMS Meteorological Data Processing System
COLL Collocation module
COMS Communication, Ocean, and Meteorological Satellite
CTP Cloud Top Pressure
GTS Global Telecommunication System
IPCC Intergovernmental Panel on Climate Change
ISCCP International Satellite Cloud Climatology Project
KMA Korea Meteorological Administration
MOD06 MODIS Cloud product
MODIS Moderate Resolution Imaging Spectroradiometer
RMSE Root Mean Square Error
SEVIRI Spinning Enhanced Visible and Infrared Imager
SGP Southern Great Plains
TSI Total Sky Imager
VAM Validation Module
Algorithm Theoretical
Basis Document
For Cloud Amount
Code:NMSC/SCI/ATBD/CA
Issue:1.0 Date:2012.12.26
File: CA-ATBD_V4.0.hwp
Page : 1/22
National Meteorological Satellite Center
1. Overview
Cloud amount (CA) is a meteorological factor which deeply influences our daily life in areas such
as sailing, aviation, agriculture, and outdoor activities. Cloudiness in an atmosphere adjusts the
intensity of incoming solar energy. This in turn regulates the global climate system. It is also a key
parameter in adjusting the emission for terrestrial radiation at the top of the atmosphere. CA is one of
the essential objects of weather forecasting, so it has long been derived by ground measurements. Due
to the importance of CA in daily life over the Korean Peninsula, it has become an object of attention to
the general public. This includes the CA of surrounding countries and bodies of water (Oh et al. 2005).
In order to acquire high quality data, The digitized retrieval of ground observation CA is requested
through satellites because of the limitations of ground observation. This section will explain the CA
algorithm for satellite observation CA and eye measurement CA including the steps and processes
involved.
2. Background and purpose
The algorithm extracts CA viewed by satellites using pixel information for the existence or
nonexistence of clouds retrieved by scene analysis. When compared with other satellites, satellite
observation CA extracts data through averaging Moderate Resolution Imaging Spectroradiometer
(MODIS) 5x5 data, Spinning Enhanced Visible and Infrared Imager (SEVIRI) 3x3 data, and (IPCC)
1x1 data. Satellite observation CA is unaffected by the location and size of clouds, but these factors
significantly affect ground observation CA. Therefore, to retrieve both satellite observation CA from
COMS Meteorological Data Processing System (CMDPS) and ground observation CA generates the
best data. CA is an essential part of the equation to predict the maximum and minimum temperature.
However, this currently relies on the ground CA observations of human observers. It has the
limitations of objective reliability with time restrictions.
Satellite observation CA from Communication, Ocean, and Meteorological Satellite (COMS) and
ground observation CA can contribute to quantitative weather forecasting. It can also be utilized in a
digital forecast by providing information with specific number and graphic form in time and space.
This product can also be utilized for the study of an indirect impact of aerosol, along with effective
Algorithm Theoretical
Basis Document
For Cloud Amount
Code:NMSC/SCI/ATBD/CA
Issue:1.0 Date:2012.12.26
File: CA-ATBD_V4.0.hwp
Page : 1/22
National Meteorological Satellite Center
particle radius of clouds.
3. Algorithm
3.1 Theoretical background and basis
Ground-observed CA is to be retrieved using inversion of method proposed by Kassianov et al.
(2005). Figure 1 illustrates the difference between ground-observed CA and satellite observation CA.
Figures 1(a) and (b) show real cloud distribution. Figures 1(c) and (d) show what can be observed
from the ground.
When many clouds exist at the zenith of ground observations without many clouds around them,
ground-observed CA can be overestimated. While clouds do not exist at the zenith of ground
observations with many clouds around them, ground-observed CA can be underestimated.
The differences between satellite-observed CA and ground-observed CA are caused by the
following:
First, there is a difference of observation conditions. Satellite observations detect the upper level
of clouds from outside of the Earth's atmosphere. Ground-observed CA detects the lower level of
clouds from the ground. Also, Satellite data has the value of pixels, but ground-observed data must
consider the curvature of the earth.
Second, ground-observed CA is affected by the horizontal and vertical structures of clouds, while
satellite observation does not consider these factors. To retrieve satellite-observed data, this
difference must be corrected. Therefore, we need to convert the results of satellite observations to
ground-observed CA. The physical law can be used to convert by the method shown in Kassianov et
al. (2005) and Oh et al. (2006).
Algorithm Theoretical
Basis Document
For Cloud Amount
Code:NMSC/SCI/ATBD/CA
Issue:1.0 Date:2012.12.26
File: CA-ATBD_V4.0.hwp
Page : 1/22
National Meteorological Satellite Center
Fig. 1. Images of (a), (b)nadir-view cloud fraction and (c), (d)fractional sky cover (Kassianov et al., 2005)
3.2 Retrieval method
3.2.1 Look up table (LUT) retrieval method
With the observer at the center, we made a look up table (LUT) for weighting values depending on
the location of clouds. The weighting value is defined by the angle ( ) between the cloud and
observer locations for maximum zenith angle in view of the observer. This weight value must be
applied to each pixel from the zenith to α=α*.For example, in the case of a cloud located over the
zenith, the weighting value becomes 0 because it does not need to consider the weighting value due to
the cloud's location. However, in the case of a cloud located around the edge, the same cloud tends to
be more overestimated than those located at the zenith. In the α* of the maximum angle increasing the
weighting value to 1.
In the case of several pixel groups being selected, the most consistent results with ground-observed
CA investigate the sensitivity and have MODIS CA and ground-observed CA of the Southern Great
Plains (SGP) region of the Atmospheric Radiation Measurement (ARM). In this process, ground
observers used a cloud base height of 2km on the assumption that they looked at cloud base.
Algorithm Theoretical
Basis Document
For Cloud Amount
Code:NMSC/SCI/ATBD/CA
Issue:1.0 Date:2012.12.26
File: CA-ATBD_V4.0.hwp
Page : 1/22
National Meteorological Satellite Center
The distance up to the edge of the clouds from the zenith of ground observers becomes about 12km,
either direction to them at the center becomes about 24km. The size of one pixel is 5km in the case of
MODIS. The edge seen by the observers needsfive pixels, one pixel in the middle, and two pixels each
direction. Thus, the observation area located at a certain point is 5x5 pixels.
When it is taken to 50km the maximum distance visible to an observer, increasing pixel size up to
15x15 and calculated for correlation coefficient and Root Mean Square Error (RMSE) with ground
observation data.Considering both the time for retrieving the ground-observed CA and consistency
with ground observation data, 5x5 appears to be the most suitable.
Table 1. Correlation coefficients and RMSE of sensitivity test to the cloud weight part.
R R5×5 R7×7 R9×9 R11×11 R13×13 15×15
SGP Correlation 0.94 0.94 0.93 0.92 0.90 0.90
RMSE 13.34 13.57 14.00 14.88 16.16 17.25
According to the sensitivity results for cloud weight value, COMS is 3km per pixel. The lookup
table to apply the weight value due to cloud position is a pixel group of 7x7 for ground-observed CA.
The retrieval method is as follows:
program makelut
!-------------------------------------------
!!Description :
! This subroutine is matrix to give cloud weight,
! == > Calculate matrix
! from formula of EVGUENI KASSIANOV et. al.
! IF fof Maximum degree(80)=distance (D) and a degree=d
! tan a = d/H
! # according to resolution, H is changed
! because tan80 = D/H, D is related to resolution
! Thus, a=tan-1(d/H)
Algorithm Theoretical
Basis Document
For Cloud Amount
Code:NMSC/SCI/ATBD/CA
Issue:1.0 Date:2012.12.26
File: CA-ATBD_V4.0.hwp
Page : 1/22
National Meteorological Satellite Center
! cloud weight= a/80
!-------------------------------------------
implicit none
integer, parameter :: grid=3, pi=3.1452, res=3
real, parameter :: H=1.6
integer :: m, n
real, dimension(7,7) :: mata, matrix
open (1, file='lut_matrix.txt', status='unknown')
DO n= 1, grid*2+1 ; DO m= 1, grid*2+1
mata(m, n)= (grid+1-m)**2 + (grid+1-n)**2
mata(m,n)= ((mata(m,n))*(res))**(0.5)/H
matrix(m,n)= ATAN(mata(m,n))
matrix(m,n)= ((matrix(m,n)) * 180.) / pi
matrix(m,n)= matrix(m,n)/80.
IF (matrix(m,n)>=1.) THEN
matrix(m,n) = 1.
ELSEIF (matrix(m,n)<1.) THEN
matrix(m,n) = matrix(m,n)
END IF
END DO ; END DO
write (1,*) matrix
end program makelut
3.2.2 Methodology
This algorithm retrieves satellite-observed CA of an interim product that converts from % unit to
ratio including pixels of clouds from the results of cloud detection. In other words, cloud pixels under
confidence level of some clear are 1. After summing up to convert into 0 in clear pixels will divide
Algorithm Theoretical
Basis Document
For Cloud Amount
Code:NMSC/SCI/ATBD/CA
Issue:1.0 Date:2012.12.26
File: CA-ATBD_V4.0.hwp
Page : 1/22
National Meteorological Satellite Center
into the number of total pixels.
The weighted value of ground-observed CA retrieved for position information of clouds from the
results of cloud detection and correcting differences that occur due to the geometric structure of clouds.
Fig. 2.Illustration of the parameters.
First, the correction for cloud position information is performed through the following process:In
the case of clouds located over the zenith, the weight value need not be considered according to cloud
position. The weight value of clouds are 0. However in the case of clouds located at theedge(i.e.,
α=α*= 80°), the same clouds tended to be overestimated than the zenith and the weight value becomes
1. Therefore, in cases of clouds locatedbetween the zenith and edge, the weight value isbetween 0 and
1.
After deciding the presence of clouds based on cloud detection information for basic input into the
algorithm, the cloud is assigned the weight value depending on ground observers and the position of
cloud based on the pixels.
In order to obtain the weight value depending on geometric information of clouds, it considers the
horizontal area and vertical height of clouds through the following process: The CA(Nnadir(α))
observed from satellites is defined as equation (1) for horizontal area that is occupied by clouds for the
area of a circle with a radius of R(α)(Figure 2). Ground-observedCA(Nhemisph(α)) is dependent upon
the zenith angle from the area occupied by the cloud for the solid angle.
Algorithm Theoretical
Basis Document
For Cloud Amount
Code:NMSC/SCI/ATBD/CA
Issue:1.0 Date:2012.12.26
File: CA-ATBD_V4.0.hwp
Page : 1/22
National Meteorological Satellite Center
(%) (1)
(%) (2)
Where Scld,nadir(α) = horizontal cloud area (km), Snadir(α) = total area of the circle (km), Scld,hemisph(α) =
fraction of the solid angle filled by clouds (sr), Shemisph
(α) = observed solid angle with cone zenith
angle 2α(sr) are defined. The relationship between CA and ground-observed CA can be expressed as
equation (3).
(3)
Where = cloud weight, = cloud aspect ratio, and , in this algorithm the observer
supposed 80°to be the maximum zenith angle.
In order to retrieve the cloud aspect ratio the horizontal and vertical size must be considered.First
of all, the horizontal area(D) is calculated considering Nnadir
Cloud type is based on ISCCP, which is divided into low, medium, and high clouds depending on
the CTP. This criteria is listed in Table 2. Bottom cloud height is determined depending on the criteria
listed in Table 2. The cloud geometrical thickness is obtained by subtracting the cloud bottom height
from the cloud top height. Thus, in low, medium, and high clouds, median CTP pressures of 300, 500,
and 850 hPa are applied, respectively. The geometrical thickness for horizontal area of the retrieved
cloud is defined by the cloud aspect ratio.
retrieved into pixel unit and area of pixels.
When considering the vertical height (H) of clouds, after dividing into low/medium/high clouds,
Cloud Top Pressure (CTP) is used to determine the geometric thickness.
The algorithm consists of three parts retrieved by ground observed CA: Weighting value of the
cloud, cloud-observed CA, and cloud aspect ratio.Ground observed CA is retrieved in cloud/clear
conditions of averaging the surrounding 7x7 pixels converting by the method of Kassianov et al.
Algorithm Theoretical
Basis Document
For Cloud Amount
Code:NMSC/SCI/ATBD/CA
Issue:1.0 Date:2012.12.26
File: CA-ATBD_V4.0.hwp
Page : 1/22
National Meteorological Satellite Center
(2005). More detail contents can be found in pages 2-3 of the appendix.
Table 2. Cloud top pressure and cloud bottom height corresponding to cloud type.
Level Genera Cloud top pressure
(hPa) Cloud bottom height
(km)
High Cirrus, cirrocumulus, cirrostratus 50-440 8
Middle Altocumulus, altostratus, nimbostratus
440-680 4
Low Cumulus, stratocumulus, stratus cumulonimbus
680-1000 1
3.3 Retrieval process
3.3.1 CA retrieval
The retrieval process of this algorithm is presented in Figure 3. The necessary input data to
estimate Satellite-observed CA and ground-observed CA are divided into basic input data and cloud
analysis data.
The basic input data is the weighting value information of clouds depending on cloud detection,
scene analysis data, and the position of the observer.
Cloud analysis data needs CTP data.
After deciding the presence of clouds based on cloud detection information, the cloud is assigned
the weight value depending on ground observers and the position of cloud by pixel detection. CA of
satellite (Nnadir
After inputting the weighted value of the clouds, cloud-observed CA, and cloud aspect ratio,
ground-observed CA is retrieved.The final product for ground-observed CA of cloud/clear condition
was generated by averaging the surrounding 7x7 pixels and converting by the Kassianov et al. (2005)
method.
) is calculated on the average cloud/clear pixel of the surrounding 7x7. We calculate the
cloud aspect ratio(γ) considering the horizontal and vertical size of cloud.
Algorithm Theoretical
Basis Document
For Cloud Amount
Code:NMSC/SCI/ATBD/CA
Issue:1.0 Date:2012.12.26
File: CA-ATBD_V4.0.hwp
Page : 1/22
National Meteorological Satellite Center
Fig. 3. Flowchart of Cloud Fraction (CF) and Cloud Amount (CA) algorithm
In order to determine satellite-observed CA, the test deciding clear sky confidence level including
scene analysis has to be preceded. If clear confidence level is too high, the number of pixels classified
according to clear pixel aredecreased and the CA is increased.If clear confidence level is too low, the
number of pixels classified according to clear pixel are increased and the CA is a decreased. The case
analysis is compared with visible image data from the COMS algorithm experiment test in order to
decide a suitable clear confidence level need. The confidence level will be within 95%.
3.3.2. QC Flag
The position information of clouds and the horizontal and vertical size of clouds influence
retrievalof ground-observed CA. Therefore, it presents each QC flag for two processes, which is
shown in table 3. First, it presents up to 96-240 for the ratio of weight value (i.e., weight_rate=sum of
cloudy pixel weight / total weight in 7x7) with the flag for position information of cloud which
derived the cloud pixels for total weighting value in the 7x7 matrix. It is designed to provide with QC
up to 1-10 for cloud aspect ratio(H/D).
Algorithm Theoretical
Basis Document
For Cloud Amount
Code:NMSC/SCI/ATBD/CA
Issue:1.0 Date:2012.12.26
File: CA-ATBD_V4.0.hwp
Page : 1/22
National Meteorological Satellite Center
Table 3.QC Flag.
CLA - CA
bit Bit Interpretation Field Description
8~5 (Pixel weights in terms of the cone zenith angle.)
unavail => 0
240 224 208 192 176 160 144 128 112 96
0~0.1 0.1~0.2 0.2~0.3 0.3~0.4 0.4~0.5 0.5~0.6 0.6~0.7 0.7~0.8 0.8~0.9 0.9~1
4~1 (Estimated Cloud aspect ratio)
unavail => 0
10 9 8 7 6 5 4 3 2 1
0~0.1 0.1~0.2 0.2~0.3 0.3~0.4 0.4~0.5 0.5~0.6 0.6~0.7 0.7~0.8 0.8~0.9 0.9~1
3.4 Validation
3.4.1 Validation method
Satellite observation CA and ground-observed CA retrieved from COMS are performed in two
ways. First is operational validation by CMDPS.Second, it has itsown validation to see the
effectiveness of the algorithm This method needs the averaging process 30kmx30km(6x6 pixels) to
adjust the spatial resolution.
In the case of ground-observed CA, after it adjusts time information between location information
of ground-observed CA and of satellite observation CA, validating the effectiveness of product on R,
Bias, and RMSE between the two products.
Algorithm Theoretical
Basis Document
For Cloud Amount
Code:NMSC/SCI/ATBD/CA
Issue:1.0 Date:2012.12.26
File: CA-ATBD_V4.0.hwp
Page : 1/22
National Meteorological Satellite Center
3.4.2 Validation data
(1) CMDPS Validation (COLL/VAM)
The data used to validate CMDPS satellite-observed used the data of CA MODIS Terra and Aqua
and ground-observed CA validated from November 1 to November 5 using Ground and GTS data. To
determine the weakness of the algorithm, we calculated the statistic (Correlation, BIAS, RMSE)
value dividing into each latitude (the equator : below latitude 30°, middle latitude: the South-North:
30-60°). Also, in CA, unlike in other products, the validation performed included clear sky.
(2) Developer validation
In order to validate satellite observation CA, MODIS cloud product (MOD06) with spatial
resolution(270x406 pixels) of 5x5 km is needed. In the case of ground-observed CA, we used ground
observation data in the SGP of the Atmospheric Radiation Measurement (ARM) program (available online
at http://www.arm.gov). In the case of ARM data, the total sky imager(TSI) was used for measuring
sky cover. It can derive more objective ground observation CA. We validated the effectiveness based
on data from August to December 2002 in the SGP site located mid-latitude similar to the location of
our country.
We used it to validate ground-observed CA at KMA, in Korea for a one month period in December
2007.
3.4.3 Temporal and spatial collocation method
(1) CMDPS Validation (COLL/VAM)
To validate thesatellite observation CA algorithm, we excluded the validation in high latitude
regions above 60° north and south.
We excluded the validation that is represented the difference more than 1-standard deviation of 5x5
pixels in MODIS to validate in the case of homogeneous position for temporal and spatial collocation
Ground data signifies the ground-observed CA at KMA. GTS are compared with cloud amount
entering within ±0.1°the observed data every hour on the hour from ground observation data. In the
case of the ground-observed CA, data more than 20 % are assumed to be navigation and cloud
detection errors and are compared with pixels within 20% for line equivalent of 1:1.
Algorithm Theoretical
Basis Document
For Cloud Amount
Code:NMSC/SCI/ATBD/CA
Issue:1.0 Date:2012.12.26
File: CA-ATBD_V4.0.hwp
Page : 1/22
National Meteorological Satellite Center
(2) Developer validation
We corresponded with temporal and spatial collocation to validate ground-observed CA retrieved
through this algorithm and ground satellite CA in the SGP site. We compared average ground-
observed CA retrieved through satellite observation CA and the algorithm to enter within ±30 minute,
±0.1 degreefrom the location of ground observation.
3.4.4 Validation result analysis
(1) CMDPS validation (COLL/VAM)
Table 4. Validation results of CF and CA
Reference Time Region R Bias RMSE
CF
MOD
11/1~11/5
Global 0.85 -3.57 21.68
MYD 0.81 -3.83 22.30
MOD Low
0.85 -5.50 21.48
MYD 0.81 -5.84 23.56
MOD Mid
0.77 -1.52 21.68
MYD 0.78 -1.86 20.90
CA
GTS
11/1~11/5
Global 0.95 0.18 0.81
GROUND 0.75 0.61 1.95
GTS Low
0.95 0.23 0.83
GROUND - - -
GTS Mid
0.91 0.19 0.78
GROUND 0.75 0.61 1.95
Table 4 shows the results in comparison with CA derived from satellite observation CA of MODIS,
Ground, and GTS. First, it shows high correlation coefficients over 0.8 in all regions if we look at the
results of satellite observation CA. When we look at the results of validation by latitude, the difference
of latitude did not influence the results of validation and it was the same as the ground-observed CA.
Algorithm Theoretical
Basis Document
For Cloud Amount
Code:NMSC/SCI/ATBD/CA
Issue:1.0 Date:2012.12.26
File: CA-ATBD_V4.0.hwp
Page : 1/22
National Meteorological Satellite Center
The results show better correlation coefficients for GTS observation results than ground. In the
case of ground, as the results observed by ground-observed CA, the objective reliability must be given
attention to decline further than GTS.
(2) Developer validation
Figure 4.shows the relationship between ground observation CA of SGP site and satellite
observation CA derived from MODIS(a), and CA converted through the algorithm, respectively.
According to Berendes et al.(2004), a disparity of within ±20 % represents “good agreement" between
Satellite and Ground observation CA.
As shown in Figure 4, the results show a correlation coefficients of 0.96 between ground and
satellite observation CA within ±20 % and represent a correlation coefficients 0.98 of Ground-
observed CA converted through ground observation CA and the algorithm.
When the overlapped large circles in Figure 4(b) compared with ground observation CA, the
percentage of the improved case for CA retrieved through the algorithm was found to be 53%.
Fig. 4. Validation results of cloud fraction(a) and sky cover(b) for the SGP site. Larger circles in b represent
improvedNhemisphvalues compared with Nnadir
in the site.
Figure 5 shows the difference between CA converted for ground and satellite observation CA, and
ground observation CA and algorithm. The result is a relatively underestimated CA in ground
observation. Thin clouds are widely distributed in mid-latitude region. ground-observed CA at the
Algorithm Theoretical
Basis Document
For Cloud Amount
Code:NMSC/SCI/ATBD/CA
Issue:1.0 Date:2012.12.26
File: CA-ATBD_V4.0.hwp
Page : 1/22
National Meteorological Satellite Center
ground seem to not exactly measure for thin cirrus.To convert ground-observed CA for algorithm from
results of Fig. 4 and Fig. 5, the correlation coefficients with ground observation CA is high and the
difference is decreased with real ground observation CA. CA retrieved through this algorithm and
ground-observed CA for one month, December at KMA, the correlation coefficient and bias are 0.61,
0.57, and RMSE denotes 3.17. This results shows that this algorithm is valid on the Korean peninsula
for smaller than RMSE (0.84) in the SGP region.
Fig. 5.monthly mean bias between ground measured cloud fraction and MODIS retrieved cloud fraction
(algorithm-retrieved sky cover).
4. Analysis method of retrieval results
The CA indicates the percentage of precision and accuracy are 10.
Table 5.Detailed Output data for the CA algorithm.
OUTPUT DATA
Parameter Mnemonic Units Min Max Prec Acc Res To
Mean cloud vertical size mean_cld_vert pixel MCV
Mean cloud horizontal size mean_cloud_hori pixel MCH
Cloud aspect ratio cld_asp_ratio 0 10 CASR
Ratio of cone zenith angles pos_weight 0 1 RA
Cloud Fraction Cloud_fraction % 0 100 pixel CF
Algorithm Theoretical
Basis Document
For Cloud Amount
Code:NMSC/SCI/ATBD/CA
Issue:1.0 Date:2012.12.26
File: CA-ATBD_V4.0.hwp
Page : 1/22
National Meteorological Satellite Center
Sky cover sky_cover % 0 100 10 10 20km CA
Prec: Precision, Acc: Accuracy, Res: Resolution
5. Problems and possibilities for improvement
In this algorithm, on account of cloud detection and CTP results used as input data, the accuracy of
this algorithm depends on the accuracy of this input data. In general, in the case of ground observation,
it does not measure thin cirrus at high altitudes and can affect ground-observed CA.
Table 6.Quality test result for the CA algorithm.
Quality flag
Parameter bit Value Meaning
sky cover 7 from 0 up to100 ; step: 1 undefined
define illumination and viewing conditions
3
0 undefined
1 night
2 twilight
3 day
4 sunglint
describe the quality of the processing itself
2
0 non processed
1 good quality
2 poor quality
3 bad quality
Algorithm Theoretical
Basis Document
For Cloud Amount
Code:NMSC/SCI/ATBD/CA
Issue:1.0 Date:2012.12.26
File: CA-ATBD_V4.0.hwp
Page : 1/22
National Meteorological Satellite Center
6. References
Berenders, T. A., Berendes, D. A., Welch, R. M., IEEE, Member, Dutton, E. G., Uttal, T. and
Clothiaux, E. E., 2004: Cloud cover comparisons of the MODIS Daytime cloud mask with surface
instruments at the north slope of Alaska ARM site, IEEE Transactions on Geoscience and Remote
Sensing, 42, 2584-2593.
Kassianov, E. I., Long, C. N. and Ovtchinnikov, M. 2005: Cloud sky cover versus cloud fraction:
whole-sky simulations and observations, Journal of Applied Meteorology, 44, 86-98.
Oh, H. R., Choi, Y. S., Ho, C. H. and Ahn, M. H., 2006: Development of sky cover retrieval algorithm
for the Communication, Ocean and Meteorological Satellite (COMS) imagery, Journal of Korean
Meteorology Society, 42, 389-396.
Algorithm Theoretical
Basis Document
For Cloud Amount
Code:NMSC/SCI/ATBD/CA
Issue:1.0 Date:2012.12.26
File: CA-ATBD_V4.0.hwp
Page : 1/22
National Meteorological Satellite Center
7. Appendix
Algorithm Theoretical
Basis Document
For Cloud Amount
Code:NMSC/SCI/ATBD/CA
Issue:1.0 Date:2012.12.26
File: CA-ATBD_V4.0.hwp
Page : 1/22
National Meteorological Satellite Center
Algorithm Theoretical
Basis Document
For Cloud Amount
Code:NMSC/SCI/ATBD/CA
Issue:1.0 Date:2012.12.26
File: CA-ATBD_V4.0.hwp
Page : 1/22
National Meteorological Satellite Center
Algorithm Theoretical
Basis Document
For Cloud Amount
Code:NMSC/SCI/ATBD/CA
Issue:1.0 Date:2012.12.26
File: CA-ATBD_V4.0.hwp
Page : 1/22
National Meteorological Satellite Center
Algorithm Theoretical
Basis Document
For Cloud Amount
Code:NMSC/SCI/ATBD/CA
Issue:1.0 Date:2012.12.26
File: CA-ATBD_V4.0.hwp
Page : 1/22
National Meteorological Satellite Center
Algorithm Theoretical
Basis Document
For Cloud Amount
Code:NMSC/SCI/ATBD/CA
Issue:1.0 Date:2012.12.26
File: CA-ATBD_V4.0.hwp
Page : 1/22
National Meteorological Satellite Center
Algorithm Theoretical
Basis Document
For Cloud Amount
Code:NMSC/SCI/ATBD/CA
Issue:1.0 Date:2012.12.26
File: CA-ATBD_V4.0.hwp
Page : 1/22
National Meteorological Satellite Center
Algorithm Theoretical
Basis Document
For Cloud Amount
Code:NMSC/SCI/ATBD/CA
Issue:1.0 Date:2012.12.26
File: CA-ATBD_V4.0.hwp
Page : 1/22
National Meteorological Satellite Center